# AI agent vs chatbot The short answer: a chatbot answers your question, and an AI agent does your task. A chatbot drafts the email; an agent finds the right person, drafts the email in your voice, and holds it for you to send. Both use the same kind of model underneath. The difference is what sits around the model: tools, memory, time, and a gate before anything goes out. ## What does a chatbot do? It takes a message and returns a reply, while you watch, inside its own window. Modern chatbots are very good at explaining, summarizing, drafting, and answering from what you paste in. When the conversation ends, so does the work. If you want the reply in your inbox, you copy it there yourself. ## What does an AI agent do? It takes a goal, works out the steps, and carries them out with tools: your connected apps, a web browser, a file system, sometimes a terminal. It keeps context between sessions, keeps working after you close the window, and comes back with the result or with a question. On Krafty Bot, an agent is a Bot with a name, a role, its own memory, and a share of your account's cloud computer, and anything it writes to a connected app waits for your approval. ## Five differences that matter - **1. Tools.** A chatbot has your words. An agent has your apps and a browser, through sign-ins you control, and uses the browser when there is no integration. - **2. Time.** A chatbot works while you watch. An agent keeps working after you close the tab and can run on a schedule. - **3. Memory.** A chatbot remembers the current conversation. An agent keeps notes between sessions. On Krafty, those are the notes you ask it to save, and you can read and edit them. - **4. Steps.** A chatbot answers one message. An agent chains many steps: read, search, compare, draft, file, rank, and checks its own progress. - **5. A gate.** A chatbot can't send anything, so it needs no gate. An agent can, so it needs one. On Krafty, writes to connected apps wait for your approval unless you allowed that action for that Bot. ## The same request, two ways - **Receipts.** Chatbot: "Here is a template for logging expenses." Agent: searches Gmail for this week's receipts, saves the PDFs to Drive, logs each in your sheet, and flags the two without a category. - **A customer bug.** Chatbot: "Here are common causes of that error." Agent: follows the customer's steps on your staging site in its own browser, captures the console error and screenshots, and drafts the issue with numbered steps. - **Outbound.** Chatbot: "Here is a cold email template." Agent: finds ten accounts that match your brief, skips anyone already in your CRM, drafts a three-sentence note for each in your voice, and ranks them in a sheet. Nothing is sent. ## When is a chatbot enough? When the output is words and you are the one who will use them: a first draft, an explanation, a summary of a document you have in front of you, a brainstorm. If the work ends when the reply arrives, a chatbot is the simpler and cheaper tool. ## When do you need an agent? When the work continues after the reply: something has to be looked up in three places, filed somewhere, compared with last week, or done again every Friday. When the result should land in your inbox, your sheet, or your tracker rather than in a chat window. And when you want to hand the job off and hear back, rather than sit with it. - [What makes an agent a teammate](https://kraftybot.app/ai-teammates) - [What is an autonomous AI agent?](https://kraftybot.app/autonomous-ai-agent) - [Nine jobs with starter prompts](https://kraftybot.app/use-cases) ## FAQs ### Is ChatGPT a chatbot or an agent? Both, depending on the mode. The chat is a chatbot. Its agent mode runs tasks on a virtual computer and asks before consequential actions. The same is true of most assistants now: the label depends on whether it can act in your tools and keep working when you close the window. ### Is an AI agent just a chatbot with plugins? Plugins add tools, which is one of the five differences. An agent also plans multi-step work, keeps context between sessions, continues when you leave, and has a gate before it commits anything. Tools alone do not make a teammate. ### Does an agent cost more than a chatbot? Usually, because it runs more model calls per job and may use a computer. Krafty meters both: The free plan includes three bots, three connected apps and three routines, with AI usage and 20 minutes of computer time that renew every month. Pro is $20 a month for up to 50 bots, unlimited connected apps, up to 50 routines, and 50 hours of cloud computer time each billing period, and it starts with a 7-day free trial. ### Can I talk to a Krafty Bot like a chatbot? Yes. Each Bot has a chat, and asking it a question works like any assistant. The difference shows when you give it a job: it goes and does it, and asks before anything goes out. ## Related pages - [AI teammates](https://kraftybot.app/ai-teammates): What turns an agent into a teammate you can hand work to. - [AI agents for small business](https://kraftybot.app/ai-agents-for-business): Which jobs to hand off first and what it costs. - [How Krafty Bot works](https://kraftybot.app/how-it-works): Memory, the shared cloud computer, apps, routines, approvals. --- Updated 2026-11-01. Page: https://kraftybot.app/ai-agent-vs-chatbot